A/B test UX design-full Review of interaction design foundation article

All Articles link:
www.interactio...
A/B test Article link:
www.interactio...
Video Description:
In this clip, we're going to be looking at what to test in terms of A/B and multivariate testing. So, this is very much around experimental questions - what to ask, what to avoid asking because they're not very appropriate questions for A/B testing, and then, finally, a few more advanced or complex questions that you might consider.
And the whole issue of experimental design is quite a complex one, and we're not going to go into the theory behind that. We're really going to be focusing on practicality. So, if you are interested in the theory in a little bit of a broader discussion of the whole topic of designing with data, there is a good book of that title by Rochelle King, Elizabeth Churchill, and Caitlin Tan, which I thoroughly recommend if you do have a hankering for some of the broader backgrounds and how this all fits into the overall user experience process. A/B testing is all about *changes in behavior*; we present people with alternative designs and we look to see how much that alters their subsequent response. So, in a simple A/B case, we show them Design A, and we show them Design B. And we measure typically a completion goal, which a lot of subject areas in user experience would refer to as *conversions*.
So, signing up to a newsletter, adding an item to a shopping basket, making a donation to a charity - these are all things that are important to their respective organizations and typically for the interactive technology that we're working on - so, websites and apps, for example; so, these are the things often that we're measuring, but they're not the only things that we can measure. We can measure really straightforward stuff like time spent on a page, time spent on the site, and also bounce rates, for example - we'll be looking at some of those a bit later on. Just a reminder that because A/B testing is done very late in the day with live sites and large numbers of users, you really want to make sure that your solution is sound before you get this far. You're not going to be able to test everything that is possibly worrying you or possibly causing problems to users;
it's just too long, involved and potentially expensive in terms of user loyalty and also the amount of effort you'd have to put into it. So, we are looking at using A/B testing and multivariate testing to basically polish a solution rather than to rework it. Bear that in mind and make sure that you've done adequate testing up to this point - so, some of the early-design methods I've talked about in the past and straightforward usability testing -before you actually get as far as shipping a new project into live use. Also bear in mind that A/B testing tends to be focused on individual pages. It is possible to have multi-page tests, but it's a more complex area than we're going to be looking at in this lesson. So, experiments have research questions. They're basically the things that you're trying to answer. And because A/B testing focuses on changes in behavior, the research questions are going to be centered on defined goals and - as I've mentioned already -typically *conversions*; so: Will, as an example, move the Add button above the fold to improve sales conversions? I would imagine it would. Actually, this is something I always find people are making the mistake of getting too talkative on the first screen of the page, and the actual Buy This or Add to Basket button gets pushed further and further down until users actually don't even see it. Will a more clearly worded charitable purpose increase donations?
So, if people have a better understanding of what your charity is about or where this money is going, would that improve conversions for those users? So, both of these can be A/B tested by using goals that you almost certainly have already defined in your analytics solution. So, these are very good candidates for A/B and multivariate testing. But I'll give you some examples of bad questions too. So, obviously, I will repeat the words "Don't ask this."
when I've mentioned them because they're not meant as examples that you should be taking away. Conversely, research questions that are not directly related to improved goal completions tend not to be suitable for A/B testing. And a kind of vague question like"Will better product photos reduce questions to customer service?" - don't ask this - is the sort of thing that you simply cannot effectively test in A/B testing.
/for reading the full description please check the link/